Questions about Swiss MMG data

Marc Ryser, Raphael Morsomme and Jason Xu

Data Overview

Two data sets

  • participants (\(n=106,817\))

  • screens (\(n=540,322\))

We have a list of 9 questions, roughly in order of importance.

Q1 – unknown / other mode of detection among women with breast cancer

632+756=1388 women with breast cancer (endpoint=1) have a detection mode that is either other or unknown.

Q1bis – Gaining information about the mode of detection

The mode of detection is a key variable for women with cancer.

Can you think of any way to gain more information on the detec field from the cancer registry?

Removing 1388 out of 4545 women with breast cancer would lead to serious bias.

Q3 – High proportion of clinical cancer

Among the 4545 cancers, 906 (20%) are coded as clinical.

This is more than expected; in the US, it is only about 20%.

Q4 – Few BI-RADS of 4, 5

While there are 4545 cancers, there are only 693+1545=2238 screens with BI-RADS of 4 or 5.

In Swiss practice, is it common for women with final BI-RADS finding of concl=3 to get a biopsy?

Q5 – Many clinical cancer diagnosed shortly after in-program screen

Is it a coincidence that many women were diagnosed as clinical right after their last screen?

Example – 15 days between last screen and endpointdt

Q6 – Non-missing BI-RADS for screens not attended?

Shouldn’t BI-RADS be missing (concl=9) when participants did not attend the screen (nopartreas=1,2,3)?

Q7 – No positive screen for screen-detected cancers

Q8 – Screens after endpointdt

Q9 – Some dates have a large number of censor dates

For instance, 1612 women were censored on May 14, 2018.